I clicked on this video with knowledge only from doing computer science at GCSE level. The way you briefly explained why you did something or what something stands for helps a lot.
Thanks a lot, your description is very clear! You've told in the video about the continuation of the series - information-driven bar construction. But I could not find it on your channel. Where is it?
I really find this trick useful to use in retrospect, like exploratory data analysis and model building. But the fact is that it's not as simple to implement other kinds of data sampling in real time data. Have you ever tried it at all? Btw, thanks a lot for the good classes.
Very good content, I would like to ask, when generating tick bars, volume bars and dollar bars, is it possible to include timestamp as well? I would like to draw candlestick charts with mplfinance. Thanks!
Hi, nice video! Converting tick data into non-time bars makes a lot of sense but I wonder what your thoughts are on doing that for example for daily price data. I've looked into it before using volume bars but it did not seem particularly successful. My explanation was that volume flow for daily bars is more homogeneous and has less information content that if the underlying data has tick resolution. I wonder if there are particular use cases for this on daily or weekly resolution. I guess, given the fractal nature of market data one could argue that the same rules should apply for lower granularity data, so I'm curious if, without giving away too much, you had any experiences with that.
So in my experience dollar bars work well except for companies like NVDA who went exponential and the daily average dollar amount changes. I have played around with ways to handle this like rolling 30 day moving averages to decide thresholds and other normalizations, but it is not as good. Companies like TSLA who haven't moved so much last few years dollar bars work amazing. Anyone else run into this?
Hi, great video! I've already seen some videos from your channel and it's very instructive and straight to the point, I'm really enjoying it. I'm kind new to Quantitative Finance, so I didn't understand how to use the processed data from this video. Is it to be the input to the strategy instead of the data with timestamp? Thanks in advance.
Hey man would appreciate it if you could go through basic mm strategies like avellaneda stoikov as the intuition behind the formulas/ methodology isnt obvious for someone untrained in financial math. Cheers
Just stumbled upon your channel, great work! How did you find doing a financial mathematics masters? I am a finance grad looking into financial engineering since I realised that I enjoyed the quantitative and technical stuff the most, like option pricing and Monte Carlo simulations.
Nice one, I'm glad you're enjoying the channel. Financial Mathematics was great, mostly enjoyed applying the theory in computational finance classes. Financial math is very involved mathematics though, really I'd say this is like 60-70% of the program and implementation is 30-40%. If you prefer implementation I recommend you save your money and stick around here ;)
@@QuantPy Thanks for an awesome answer! I will definitely stick around here to play around to learn some slick simulations to run in Python. Your channel made me realize how fun and effective Python can be. When it comes to the math stuff I loved it as a kid but struggled with it until I rediscovered it with the finance papers, where I finally found a fun context with it. I am actually looking into doing some extra maths papers to prepare!
All you would need to do is capture the time stamp at the beginning/end of the bar. I have not done this in my aggregation as it was not inbuilt into ‘ohlc’ pandas function.
I recommend you watch my video on historical VaR calculations. One suggestion would be to remember to use historical log returns series for VaR calculations
Also, the book fails to address multiple problems with dollar/tick/whatever bars such as: pump and dump schemes which purposely inflate the amount of trading activity that exists (are these traders really "informed" as the book says?), and related wash trading schemes which generate a lot of fake trading activity. How can your machine learning model determine whether or not you're training on these kinds of activities generated from your dollar/tick/whatever bars?
I think both of your calculations for volume and dollar bars are wrong. Your method is equivalent to sampling based on the accumulation of volume & dollar volume // threshold, which creates problem when you have observations with large values.
I clicked on this video with knowledge only from doing computer science at GCSE level. The way you briefly explained why you did something or what something stands for helps a lot.
Thanks a lot, your description is very clear!
You've told in the video about the continuation of the series - information-driven bar construction. But I could not find it on your channel. Where is it?
Great content. I just started in the financial industry and have a long way to go. Thank you for sharing this!
Interesting video. De Prado can be very quant so anyone who can simplify his analysis would be great.
Thanks. Very useful video.
Thanks for the video on forming alternative data sets on volume and dollars. Do you have a video on trading algorithmically with this kind of data?
How do we calculate the correlation between two assets if the tick bars occur at different times?
Very useful. Tnx for sharing.
I really find this trick useful to use in retrospect, like exploratory data analysis and model building. But the fact is that it's not as simple to implement other kinds of data sampling in real time data. Have you ever tried it at all? Btw, thanks a lot for the good classes.
thank you so much for this video!!
Thank you
When will you make information driven bars?
great content! do you have any suggestion how to clean the tick data?
Very good content, I would like to ask, when generating tick bars, volume bars and dollar bars, is it possible to include timestamp as well?
I would like to draw candlestick charts with mplfinance. Thanks!
Yes
Really an underrated channel.
Which section in Commsec did you pull the intraday data?
This is great!
Should create a discord server for everyone to chat. I'm based on the Gold Coast, starting as a grad quant early next year
Hi, nice video! Converting tick data into non-time bars makes a lot of sense but I wonder what your thoughts are on doing that for example for daily price data. I've looked into it before using volume bars but it did not seem particularly successful. My explanation was that volume flow for daily bars is more homogeneous and has less information content that if the underlying data has tick resolution. I wonder if there are particular use cases for this on daily or weekly resolution.
I guess, given the fractal nature of market data one could argue that the same rules should apply for lower granularity data, so I'm curious if, without giving away too much, you had any experiences with that.
So in my experience dollar bars work well except for companies like NVDA who went exponential and the daily average dollar amount changes. I have played around with ways to handle this like rolling 30 day moving averages to decide thresholds and other normalizations, but it is not as good. Companies like TSLA who haven't moved so much last few years dollar bars work amazing. Anyone else run into this?
Hi, great video! I've already seen some videos from your channel and it's very instructive and straight to the point, I'm really enjoying it. I'm kind new to Quantitative Finance, so I didn't understand how to use the processed data from this video. Is it to be the input to the strategy instead of the data with timestamp? Thanks in advance.
Thank you.
Sir if you have created the playlist please share with us.
Hey man would appreciate it if you could go through basic mm strategies like avellaneda stoikov as the intuition behind the formulas/ methodology isnt obvious for someone untrained in financial math. Cheers
How do you get the threshold for the dollar bars?
This is at your discretion as an analyst
Just stumbled upon your channel, great work! How did you find doing a financial mathematics masters? I am a finance grad looking into financial engineering since I realised that I enjoyed the quantitative and technical stuff the most, like option pricing and Monte Carlo simulations.
Nice one, I'm glad you're enjoying the channel. Financial Mathematics was great, mostly enjoyed applying the theory in computational finance classes. Financial math is very involved mathematics though, really I'd say this is like 60-70% of the program and implementation is 30-40%. If you prefer implementation I recommend you save your money and stick around here ;)
@@QuantPy Thanks for an awesome answer!
I will definitely stick around here to play around to learn some slick simulations to run in Python. Your channel made me realize how fun and effective Python can be.
When it comes to the math stuff I loved it as a kid but struggled with it until I rediscovered it with the finance papers, where I finally found a fun context with it. I am actually looking into doing some extra maths papers to prepare!
How would one actually plot dollar bars with time?
All you would need to do is capture the time stamp at the beginning/end of the bar. I have not done this in my aggregation as it was not inbuilt into ‘ohlc’ pandas function.
Nice class, wrong take on ticks. except your csv is containing single trade executions...
Anyway, keep the videos coming ;)
Hello my frind can u calculate Historical Value at Risk (VaR) for bitcoin price? Really struggling a lot with this
I recommend you watch my video on historical VaR calculations.
One suggestion would be to remember to use historical log returns series for VaR calculations
Also, the book fails to address multiple problems with dollar/tick/whatever bars such as: pump and dump schemes which purposely inflate the amount of trading activity that exists (are these traders really "informed" as the book says?), and related wash trading schemes which generate a lot of fake trading activity. How can your machine learning model determine whether or not you're training on these kinds of activities generated from your dollar/tick/whatever bars?
Comment for the RUclips algorithm
Legend!
@@QuantPy BTW cool content.
I think both of your calculations for volume and dollar bars are wrong. Your method is equivalent to sampling based on the accumulation of volume & dollar volume // threshold, which creates problem when you have observations with large values.
Use tick charts 4000